In openejb the number of classes generated is very small (a few per
EJB deployment) as compared to Geronimo which generates tons. IIRC,
generation in geronimo accounts for over 25% of startup time. I
highly doubt that OpenEJB has this problem, but if someone has some
benchmarks that show this to be a major performance (startup) issue,
I'd be willing to look at a caching system. Until then, I'll take
the simple road.
-dain
On Dec 24, 2006, at 7:11 AM, Matt Hogstrom wrote:
Not sure how tough this would be but I've heard from some users
that dynamically generating the code at startup is really expensive
for them (lots of EJBs). They asked if we could save the generated
artifacts so they would incur the cost once.
On Dec 21, 2006, at 1:25 AM, Dain Sundstrom wrote:
I have created a bunch of subtasks for https://issues.apache.org/
jira/browse/OPENEJB-123 so you can track my progress.
Currently, I have CMP 1.x working and a good portion of CMP 2.x.
For 2.x, I haven't written the code that generates concrete
subclasses of the abstract CMP 2.x entity bean class, so you have
to code them by hand like this:
http://svn.apache.org/viewvc/incubator/openejb/trunk/openejb3/
itests/openejb-itests-beans/src/main/java/org/apache/openejb/test/
entity/cmr/onetomany/ABean.java?revision=487456&view=markup
This is the next thing I'm going to work on which means I need to
get JPA runtime enhancement working (since I'll be generating the
classes at runtime).
Anyways, other then the class generation junk, the rest of the
code base is dirt simple, and will get even simper of the next few
weeks, so if you get some free time give it a read (also the MDB
container :))
-dain
Matt Hogstrom
[EMAIL PROTECTED]
